Archal
Stateful API testing sandboxes for AI agents: no test accounts, no rate limits, with state that persists or can be reset.
About Archal
Archal provides testing sandboxes for AI agents. Each sandbox is a stateful environment whose behavior simulates the real APIs you use in production, letting you verify API integrations without registering test accounts, worrying about rate limits, or polluting shared test data. Every test run gets an isolated sandbox: changes within the environment persist across tool calls and are reset back to the same baseline state when the run ends. The website lists two main use cases: integration testing (dedicated sandboxes for each branch or CI job) and agent training with reinforcement learning (a fresh sandbox with the same starting state for every rollout, with changes persisting throughout the episode). Self-service use requires no subscription and no credit card, and billing is based on the minutes each environment runs.
Best for
Best use cases
Run isolated API integration tests for every branch or CI job
Verify third-party API integrations when you have no test accounts on hand or don't want to pollute shared test data
Provide environments with the same starting state and persistent changes within an episode for reinforcement learning or agent training
Let coding agents (such as Claude Code, Codex, Cursor, and Devin) call sandboxes directly for verification
Features
Stateful sandbox environments
Each sandbox includes a stateful environment that simulates the APIs you use in production, so you can test API integrations without test accounts.
Persistent and resettable changes
Changes within the environment persist across tool calls and are reset back to the same baseline state when the run ends.
Isolated sandboxes per branch / CI job
Every branch or CI job can get its own isolated sandbox, avoiding shared test accounts, rate limits, and dirty state.
Agent training and RL support
Provides a sandbox with the same starting state for every rollout, with changes persisting throughout the episode.
API key access
Access directory, sandbox, state, and evidence endpoints via a workspace API key; keys can be restricted to specific environments and lifecycle scopes.
Scoped credential isolation
Data-plane credentials for environments are bound to a specific workspace, session, environment, and expiry; the gateway rejects workspace keys and real upstream credentials.
Pros
- Each sandbox is a stateful environment where changes persist across tool calls and can be reset back to the baseline
- No need to register test accounts, and free from rate limits and dirty state
- Self-service use with no subscription and no credit card; you get a one-time $20 credit upon signup
- Failed provisioning and cold-start time are not billed, with finer per-second proration
- One-click pairing with common coding agents such as Claude Code, Codex, Cursor, and Devin
Cons
- New sandbox creation pauses once the balance is used up, and paid continuation won't be available until the end of the safe rollout, with no recovery timeline stated on the website
Pricing Plans
Self-service pay-as-you-go
- Receive a one-time $5 credit after signup verification, plus another $15 once the first sandbox is ready
- Each active environment is billed at $0.10 per minute, prorated by the second
- Failed provisioning and cold-start time are free
- No credit card required, no subscription
* Pricing information is for reference only. Please refer to the latest announcement on Archal's official website.